DeFiPP Research Seminar-Gerdie Everaert (Ghent University)

Bias of the Fixed Effects Estimator in Panel Data Local Projections

Abstract

Local projections (LPs) have become a popular tool for the estimation of impulse response functions (IRFs). Compared to traditional dynamic models, LPs allow for more flexible IRF estimation by imposing weaker assumptions on the dynamics of the data. Since LPs are a set of linear dynamic models, they can be consistently estimated with least squares methods as the number of time periods T grows to infinity. However, not much is known about the bias for finite T. This is especially relevant in a panel data setting, where the number of individuals N tends to infinity and T is kept fixed. In this paper, we derive an analytical bias expression for the LP fixed effects estimator. We show that the bias can be sizeable, even for moderately large values of T. We further show that a GLS transformation in combination with a two-tier bias correction eliminates most of the bias. We illustrate the practical relevance of our approach by estimating the impact of commodity price shocks on CPI inflation in 144 countries from January 2000 to December 2010. Compared to the FE LP estimates, the bias-corrected GLS LP estimator suggests a much larger impact, which is almost three times as large 48 months after the shock.

Contact : Romain Houssa - romain.houssa@unamur.be
Télecharger : vCal